February 1, 1996
fficials at the NASA Aerospace Simulation (NAS) Facility first began using the most powerful computers (called high-performance or supercomputers) a decade ago with the goal of reducing, if not eliminating, wind tunnel testing of new aircraft. Wind tunnels, while extremely effective at evaluating how an aircraft will perform in flight, were expensive. Because tests took years to complete, major U.S. aircraft companies were unable to get their products to market fast enough to compete effectively in the global economy. NAS officials, who routinely perform long-term research and development for industry, believed supercomputers and high-end applications could recreate an airplane on screen and simulate factors such as turbulence and air flow.
Unfortunately, because of the cost and complexity of advanced computing, NAS officials struggled to advance their goals. But then in 1991, the National High Performance Computing and Communications (HPCC) program was created. Several agencies joined up and began exchanging data and resources, enabling NAS officials to collaborate with federal scientists and engineers working on climate control simulation and other related models.
"Previously, we were trying to solve some specific system software problems but we could only do so much considering the cost of the machines that we needed and the amount of money that we had," says William Feireisen, project manager for computational aerosciences for the High Performance Computing and Communications Program at NASA's Ames Space Center. "With the HPCC program, we get a chance now to look at all the other system software work that's being done and we can leverage that."
Today, thanks to this cooperative effort, people in the NAS program have increased their supercomputing capabilities by a thousand-fold. They can not only simulate aircraft using three-dimensional models, but they can link up with aircraft companies via high-speed networks, enabling industry, university and agency experts to collaborate with NAS officials on improving the simulation process. Although the work continues, one early result has been a marked improvement in the time it takes a new product to reach the market.
"What we're really doing is developing the kind of tools that these companies will need to create better, safer, more environmentally sound planes and also to compete better economically," says Walter Brooks, division manager of the Advanced Computing and Communications Technology Division at Ames. "We're creating a general competitive advantage or base from which all of the U.S. industry can launch."
The HPCC program acts as a "virtual agency," helping to coordinate the ongoing efforts of 12 federal agencies-as well as academic scientists and industry researchers-to develop and improve supercomputing and advanced communications technologies like networking, virtual reality, digital libraries, electronic commerce and scientific visualization. "I think it's pretty well recognized that the same needs for high-performance computing exist all the way across the government," says Feireisen, "and that these needs are so large and so difficult that an individual agency just does not have the horsepower to address them as effectively as a cooperative effort does."
The goals of the program include the following:
1. Increase the productivity and competitiveness of key U.S. industries by addressing fundamental problems in science and engineering, such as developing computationally intensive environmental models to ensure accurate monitoring of acid rain and ozone levels.
2. Provide key support for the National Information Infrastructure (NII), a White House program that enhances access to government information through advanced high-speed networks and software. The NII application video-on-demand, for example, will rely on scalable parallel computing and mass storage systems developed by high-performance computing research centers.
3. Allow scientists and other researchers to collaborate on projects through robust networks.
4. Perform long-term research to safeguard the United States' status as a leader in high-performance computing and networking technologies.
Like other federal programs, HPCC will likely take hits in the budget battles on Capitol Hill, but not because of any major policy change. Since 1991, the HPCC program has received significant bipartisan support.
The HPCC program comprises five components: High Performance Computing Systems, the National Research and Education Network, Advanced Software Technology and Algorithms, Information Infrastructure Technology and Applications, and Basic Research and Human Resources. Each component is equally important, for if one does not advance at the same pace, then the remaining four are also halted in their progress. Most people agree that systems software is the one area slowing up progress. "There has to be a balance," says John C. Toole, director of the HPCC's National Coordinating Office. "I think in general, software will always lag. It needs to be tightly coupled with the best and brightest people, but it takes time to emerge from a research point of view."
Agencies participating in the HPCC program include the Defense Advanced Research Projects Agency, the National Institutes of Health, the National Science Foundation, the National Oceanic and Atmospheric Administration, the Environmental Protection Agency, the Education Department and the National Institute of Standards and Technology.
As in the NASA Aerospace Simulation program, nearly all federal offices involved in HPCC began investing in supercomputing research and development long before the program was created by legislation. For example, the Internet, a worldwide network of networks that today connects more than 4 million computers, was set in motion nearly 30 years ago when DARPA devoted high-performance computing resources to develop a network that could connect researchers.
The HPCC cross-agency approach has eliminated much duplication in effort and equipment, created opportunities to collaborate and share lessons learned, and gained considerable support and assistance from information technology vendors. For example, many companies today lease, rather than sell, high-end machines to agencies. And in response to agency needs, Digital Equipment Corp. and Cray Research Inc., recently developed methods of linking machines together, an advancement that will make high-end computing more affordable.
"You really get the industry's attention when you've got a national program like this," says Feireisen. "One of the key points of this is it allows all of the agencies to speak with a uniform voice so the common problems are communicated to all the various computer vendors."
Toole adds, "I think we have much earlier access to a very vibrant research community, so that the participating agencies get a much earlier handle on what's real and what isn't, what's evolving and what can be appropriately applied to their own organization."
Such advantages have spurred agencies to take on ambitious projects. HPCC is allowing "an extraordinary increase in people's ability to deal with real problems," says Michael Levine, scientific director of the Pittsburgh Supercomputing Center. "Not problems of simply academic interest, but real achievement in such fields as health care, the environment and transportation."
Toole notes that the HPCC's goals have already been well-integrated with agency goals and programs. But most agree that while the infrastructure is in place and advances have been made, research and development efforts have only just begun. NASA, for instance, believes that it needs to increase its computing ability by another 1,000 percent before it will fully achieve its flight simulation goals.
"We're advancing on all fronts," says Levine. "As in any complicated endeavor there are fits and spurts and certain areas are tougher than others. But it is progressing at a rate that I never would have guessed was possible five years ago."
February 1, 1996